2011, Krüger, Fabian, Mokinski, Frieder, Pohlmeier, Winfried

This paper reinterprets Maganelli's (2009) idea of "Forecasting with Judgment" to obtain a dynamic algorithm for combining survey expectations data and time series models for macroeconomic forecasting. Existing combination approaches typically obtain combined forecasts by linearly weighting individual forecasts. The approach presented here instead uses survey forecasts in the estimation stage of a time series model. Thus an estimate of the model parameters is obtained that reflects two sources of information: a history of realizations of the variables that are involved in the time series model and survey expectations on the future course of the variable that is to be forecast. The idea at the estimation stage is to shrink parameter estimates towards values that are compatible (in an appropriate sense) with the survey forecasts that have been observed. It is exemplified how this approach can be applied to different autoregressive time series models. In an empirical application, the approach is used to forecast the three-month Euribor at a six-month horizon.

2010, Krüger, Fabian, Mokinski, Frieder, Pohlmeier, Winfried

This paper reinterprets Maganelli’s (2009) idea of “Forecasting with Judgment” to obtain a dynamic algorithm for combining survey expectations data and time series models for macroeconomic forecasting. Existing combination approaches typically obtain combined forecasts by linearly weighting individual forecasts. The approach presented here instead uses survey forecasts in the estimation stage of a time series model. Thus an estimate of the model parameters is obtained that reflects two sources of information: a history of realizations of the variables that are involved in the time series model and survey expectations on the future course of the variable that is to be forecast. The idea at the estimation stage is to shrink parameter estimates towards values that are compatible (in an appropriate sense) with the survey forecasts that have been observed. It is exemplified how this approach can be applied to different autoregressive time series models. In an empirical application, the approach is used to forecast the three-month Euribor at a six-month horizon.